The study also found that ChatGPT only answered math problems correctly 50% of the time, and its step-by-step problem-solving approach was incorrect 42% of the time. The researchers likened the issue to an overreliance on autopilot, suggesting that while students might answer more problems correctly with ChatGPT, they actually learn less. The study, which involved nearly a thousand students, has not yet been peer-reviewed and more research is needed to confirm its findings.
Key takeaways:
- Researchers at the University of Pennsylvania found that Turkish high school students who used AI chatbot ChatGPT for math practice performed worse on a math test compared to those who didn't use the chatbot.
- Even a revised version of ChatGPT that functioned more like a tutor didn't improve test scores, suggesting that students were using the chatbot as a crutch and not building problem-solving skills.
- ChatGPT only answered math problems correctly half of the time, with its step-by-step approach for solving a problem being wrong 42 percent of the time.
- The researchers likened the problem of learning with ChatGPT to autopilot, highlighting the risk of overreliance on technology in education.